Question answering systems typically utilize a large set of documents, referred to as corpora, to perform searches and answer questions. As part of the search and question answering process, question answering systems acquire evidence from the corpora to generate hypotheses and identify further evidence to support the hypotheses. To improve accuracy, question answering systems typically utilize domain-specific corpora that map to specific vocabularies within a specific domain.
The evidence-gathering process may discover patterns of information in the corpora, such as detecting particular word patterns over time. The patterns that are discovered, however, are typically the most apparent patterns and typically within a particular document. As a result, less-apparent patterns of evidence that spans across corpora may go undiscovered.